Research On The Recognition Methods Of Paper-Cutting

Image recognition is one of the hot issues and has its wide application in every field of society, more and more experts and scholars pay attention to the development of image recognition, and have achieved very good results in this area. Paper-cutting is a traditional folk art with a long history in China. With the development of animation industry, paper-cutting becomes a good cartoon material. As works of art will deform easily, image recognition becomes more difficult. Nowadays there are few studies in this area, so research on computer paper-cutting that combines paper-cutting and image recognition would be an important and significant work.Feature extraction is the key to decide similarity and identify images problem and how to extract invariant features from images is a core problem. This paper, based on the research of feature extraction methods, recognition methods,current domestic and international research status, propose some new methods that improved the non-mathematical transformation. Theoretical analysis and experimental results have proved that the proposed methods are feasible and effective. The application of these methods in paper-cutting image is also completed.The main works in the thesis are as follows:(1)Research on the paper-cutting image's characteristic, and sum the types of the paper-cutting patterns. Pre-processing is an important task in image processing. In the first place, template method is used to effectively de-noise the background noise of scanning images, and then image graying by the weighted average method, finally change gray images into binary images to prepare for the follow works.(2)Considering a simple way to extract image features, we propose an R- transform and singular value decomposition (SVD) feature extraction algorithm based on the radon transform. The singular value feature of the image after the R-transform represents the image structural feature, and has a strong robustness. The experiment proves that this algorithm is invariant to rotation, scale and translation as well as suitable for the recognition of paper-cutting patterns.(3)For the shortcomings of existing feature extraction methods is not suitable to the morph image, a new method is proposed in this paper. Based on the invariance of Fourier-Mellin transform, calculate the energies of the subbands acquired by wavelet transform, and then the invariant feature vector that composed of the energies. Experiments indicate that this feature vector not only has the translation, rotation and scale invariance, but also satisfactorily achieves the pattern recognition.(4)This thesis researches on various image recognition methods. As regards the Support vector machine (SVM), it has a superior capability for generalization and this capability is independent of the dimensionality of the input data. In this study, SVM is used to classify the paper-cutting pattern classification. The simulation results show that SVM is a good method for pattern classification.The images of paper-cutting used in the experiment are all scanning copies of some related books; some theoretic analysis and verification toward the arithmetic mentioned in the thesis are done through experiments, its results show:(1)The analytical method based on R-transform and singular value decomposition, it has the simple computation, good characteristic robustness, translation, rotation, scale invariance and better differentiation toward most deformed images.(2)Multiresolution FM transform algorithm proposed is invariant to similarity transformation, producing accuracy and effective recognition of paper-cutting patterns.(3)SVM has proven to be a powerful technique for pattern classification. It is an effective and general method for representing complex function in high dimensional space. The experiments results show that SVM is a good method for paper-cutting patterns recognition.Based on the practical problem, this theist has deeply researched on the image pre-processing, feature extraction and image recognition methods etc. Aiming at the non-mathematical deformation image, it proposes some useful methods in feature extraction and image recognition.